Design Of Crop Disease And Insect Pest Identification System Based On CNN

Crop image classification is one of the urgent problems to be solved in agricultural development.The ability to identify normal crops quickly and accurately in visible images is an important means to realize smart agriculture today.This article puts forward a method for identifying and classifying crop leaves based on Convolutional Neural Network(CNN)targeting the classification of most visible images of leaves.Based on the improvement of the classic RESNET50 model,this method combines with Center loss and Softmax loss,thereby reducing the loss function of CNN and optimizing the robustness of the model.Therefore,the feature learning effect and classification accuracy are successfully improved,leading to a great leap in the image recognition of crop leaf diseases and insect pests.This article first samples and collects the data needed for the experiment,and then uses image segmentation technology to preprocess the sampled images.Next,it introduces in detail the various network models,loss functions,and optimization algorithms applied by this method.Based on the above,it completes the overall design and coding of the system.Finally,the experiments are tested,which proves the effectiveness of this algorithm for network optimization.Compared with the traditional network models,the learning method proposed in this paper applies the Focal loss function and the Adam optimization algorithm to optimize the model,which is thus of more excellent classification performance and generalization ability,thereby more effectively realize the identification of crop diseases and pests.